Ad hoc parsers are pieces of code that use common string functions like split, trim, or slice to effectively perform parsing, without employing any formal parsing techniques or frameworks. The grammars of ad hoc parsers remain entirely implicit. Is it possible to infer such grammars, in the same way that we can infer types? We believe that it is and present Panini, an intermediate representation for ad hoc parsers based on refinement types.

In this extended abstract, we report our progress toward grammar inference via refinement types and highlight the challenges and work still ahead of us.